Impala是一个开源的,基于Hadoop的分析型数据库。
Impala可以查询存储在HDFS或者HBase中的数据。
Impala通过专用分布式查询引擎,绕过MapReduce直接访问数据,查询性能远高于Hive。
3 impyla
基于HiveServer2 实现的分布式查询引擎(如Impala、Hive)的Python客户端。
完全符合DB API 2.0(PEP 249)规范。
使用Kerberos、LDAP、SSL。
支持将数据转换为pandas的DataFrame,轻松集成到Python数据栈(如scikit-learn、matplotlib等)。
4 类封装
class Impala(SQL):
DESC_EXEC_SUCCESS = "执行成功"
def __init__(self, host, port, database, user, password=None):
"""Impala工具类
:param host: IP
:param port: 端口
:param database: 数据库名
:param user: 用户名
:param password: 密码
self.host = host
self.port = port
self.database = database
self.user = user
self.password = password
self.connect = None
self.cursor = None
def get_connect(self, timeout=600):
"""获取连接
:param timeout: 超时时间
self.connect = connect(
host=self.host,
port=self.port,
timeout=timeout,
database=self.database
def get_cursor(self):
"""获取游标
self.cursor = self.connect.cursor(
user=self.user
def close(self):
"""关闭连接
self.cursor.close()
self.connect.close()
self.cursor = None
self.connect = None
def execute(self, sql, auto_close=True):
"""执行sql
:param auto_close: 执行结束是否自动关闭连接
if not self.connect: self.get_connect()
if not self.cursor: self.get_cursor()
self.cursor.execute(sql)
try:
result = self.cursor.fetchall()
except ProgrammingError:
result = self.DESC_EXEC_SUCCESS
if auto_close: self.close()
return result
5 使用例子
from utils.db.impala import Impala
impala = Impala(
host="10.123.0.11",
port=123456,
database="fields",
user="unclebean"
sql = "select 1 as a, 2 as b union all select 3 as a, 4 as b"
result = impala.execute(sql, auto_close=True)
print(result)
默认的返回结果是一个列表,列表中每个元素代表一行结果,类型为元组,如上面返回的结果:[(1, 2), (3, 4)]
若想行结果变为字典,而非元组,则获取游标时需传入参数dictify=True,如
sql = "select 1 as a, 2 as b union all select 3 as a, 4 as b"
result = impala.execute(sql, auto_close=True, dictify=True)
print(result)
则返回结果变为:[{'a': 1, 'b': 2}, {'a': 3, 'b': 4}]
6 tkinter封装,实现一键刷新元数据和按日统计数据量
import os
from datetime import datetime
from menu.menu import EMenu
from utils.db.impala import Impala
class MenuImpala(EMenu):
LABEL_NAME = "Impala"
LABEL_NAME_INVALIDATE_METADATA = "Invalidate Metadata"
LABEL_NAME_COUNT_BY_DAY = "Count By Day"
DESC_SSH_CMD_FAILED = "执行错误"
def __init__(self, master=None, cnf={}, **kw):
super().__init__(master=master, cnf=cnf, **kw)
self.impala = Impala(
host = self.conf.impala.HOST,
port = self.conf.impala.PORT,
database = self.conf.impala.DATABASE,
user = self.conf.impala.USER
master.add_cascade(label=self.LABEL_NAME, menu=self)
self.add_command(
label=self.LABEL_NAME_INVALIDATE_METADATA,
command=self.invalidate_metadata
self.add_command(
label=self.LABEL_NAME_COUNT_BY_DAY,
command=self.count_by_day
@EMenu.thread_run(LABEL_NAME_COUNT_BY_DAY)
def count_by_day(self):
"""菜单命令:按日统计数据量
table_name = self.get_table_name_from_clip()
self.invalidate_table(table_name, auto_close=False)
sql = "select data_date,count(1) from {} group by data_date order by data_date desc".format(
table_name
self.stdout(sql, with_time=" - ")
result = self.impala.execute(sql)
result = "\n".join([str(row) for row in result])
self.stdout("{} -> {}".format(sql, result), with_time=" - ")
self.msg_box_info(result)
@EMenu.thread_run(LABEL_NAME_INVALIDATE_METADATA)
def invalidate_metadata(self):
"""菜单命令:刷新元数据
self.invalidate_table(self.get_table_name_from_clip())
def invalidate_table(self, table_name, auto_close=True):
sql = "invalidate metadata {}".format(table_name)
self.stdout(sql, with_time=" - ")
result = self.impala.execute(sql=sql, auto_close=auto_close)
self.stdout("{} -> {}".format(sql, result), with_time=" - ")
def get_table_name_from_clip(self):
table_name = self.paste()
if len(table_name.split(".")) == 1:
table_name = table_name.split("_")[0] + "." + table_name
return table_name
7 完整代码
GitHub上搜索TheUncleWhoGrowsBeans
数据 @ 一闪
22.0k
粉丝